How Fovus enabled a biotech to accelerate high-throughput screening by up to 96x with cloud HPC

Let’s face it: Cloud-based High-Performance Computing (HPC) offers unprecedented computational power. These systems can perform quadrillions of calculations per second. But that computational horsepower comes at a price – not just financially, but in the form of a steep learning curve and a constant battle against configuration headaches, unpredictable costs, and delays.

For researchers working in fields such as drug discovery, genomics, and materials science, cloud-based HPC has historically been, at times, both a blessing and a curse. While it can help researchers redraw the boundaries of scientific inquiry through massive-scale simulations, analysis of complex datasets, and integration of AI, the inherent complexity of managing cloud HPC infrastructure often presents a significant obstacle.

Fengbo Ren, CEO of Fovus, points out that “users are facing significant challenges in scaling up the amount of computation.” Many researchers lack the specialized knowledge required to navigate the intricacies of cloud environments. “They are domain experts,” Ren says. “But they tend to have very little cloud experience.”

Cloud complexity can be considerable

While cloud computing has streamlined many IT tasks, it introduces new challenges for researchers. These range from billing complexities and migration issues to an overwhelming array of tools and high demand for GPUs, often leading to cost overruns.

Another major challenge in cloud-based HPC is the prolonged “time-to-insight” — the time it takes to achieve meaningful results. As Ren explains, “Without good management of their cloud platform, researchers can get the time to insight up to weeks, which is really unbearable.”

Fovus aims to address these challenges with its AI-powered serverless HPC platform. By automating cloud logistics, optimizing HPC strategies based on benchmarking data, and identifying cost-effective spot instances, Fovus enables researchers to significantly accelerate their work, reduce cloud expenses, and focus on scientific discovery. Ren highlights that Fovus “helps reduce that time to insight from weeks or days to hours by optimizing the allocation of cloud resources.” This reduction empowers researchers to accelerate their research process and explore a wider range of possibilities. Ren points to a case study where Fovus reduced a research team’s monthly cloud bill from $20,000 to $5,000. Its website also has case studies involving dramatic HPC efficiency gains at Komatsu and the CRO Chemspace.

Biopharma AI demand stokes HPC boom

The pharma and biotech sectors are emerging as prominent HPC adopters, Ren noted. “The trend of AI-based drug discovery has led to an explosion in HPC demand,” he adds. Based on the company’s own customer data, Fovus estimates that HPC usage in these sectors is about 20 times higher per user than in the engineering industry.

One example of Fovus’s impact is a stealth Series-B biotech startup focused on developing innovative therapies. This startup was struggling with the complexities of managing cloud HPC for crucial tasks like high-throughput virtual screening (HTVS). Its team of computational chemists faced long wait times for results and high cloud costs. By implementing Fovus, the startup saw their HTVS workflows achieve an up to a 96x speed increase, as Fovus notes on its website. That gain involved AI-augmented HTVS — a computationally intensive process that previously took 2 days to complete. Fovus trimmed the time-to-insight to 30 minutes. In addition, the startup also experienced a substantial drop in cloud expenses, with costs for some HTVS workloads falling to one-eighth of their prior level.

An AI-enabled serverless HPC approach born from frustration

Fovus, like many startups, was born out of frustration. During the mid-2010s, while working as an associate professor at Arizona State University, Ren encountered significant challenges attempting to migrate HPC workloads to the cloud using existing technologies. These frustrations weren’t unique to him. In conversations with other industry professionals, Ren discovered that many researchers shared similar struggles with cloud HPC initiatives. “That’s when I realized there were very common pain points,” he says. “This seemed to be very urgent.” Determined to find a better solution, Ren took matters into his own hands, founding Fovus in 2021 with the mission of simplifying cloud HPC.

Fovus’ AI-powered engine optimizes resource selection, preventing overspending and maximizing utilization. It uses automation to streamline infrastructure provisioning and management. In addition, Fovus taps cost-effective spot instances while mitigating their inherent risks.

Ultimately, Ren sees Fovus as part of a much-needed shakeup in the HPC industry. “HPC is a legacy industry,” he says. “It actually existed since the advent of a computer back in the 1950s… So it was a very well isolated, traditional industry. And I think revolution has been long overdue.”

---

Filed Under: Drug Discovery, machine learning and AI
Tagged With: AI-powered HPC, biotech research acceleration, cloud computing costs, computational resources, drug development pipelines, pharma R&D efficiency, serverless computing
 

---